Improved multiplexed metropolis light transport

Zuo Zeng; Meie Fang; Lizhuang Ma; Xiao Lin

Improved multiplexed metropolis light transport

Zuo Zeng
, Meie Fang
, Lizhuang Ma
, Xiao Lin^*

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

2 Scopus citations

Abstract

Aiming at the problem that multiplexed Metropolis light transport (MMLT) may not promise coherency in path space and does not reuse sub-paths, this paper introduces an improved algorithm. It firstly samples and selects sub-paths to reuse paths. Secondly, it sorts these sub-paths to explore the path coherency in the path space. Finally, it designs a set of weighting strategies to better compute weights of different sampling techniques. Experiments show that the algorithm produces less noisy results under the same conditions.

Original language	English
Pages (from-to)	1459-1464
Number of pages	6
Journal	Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics
Volume	28
Issue number	9
State	Published - 1 Sep 2016
Externally published	Yes

Keywords

Global illumination
Markov chain Monte Carlo
Multiple importance sampling
Path reuse

Cite this

@article{3a00b4f64db1487c9aa938c0f279c125,

title = "Improved multiplexed metropolis light transport",

abstract = "Aiming at the problem that multiplexed Metropolis light transport (MMLT) may not promise coherency in path space and does not reuse sub-paths, this paper introduces an improved algorithm. It firstly samples and selects sub-paths to reuse paths. Secondly, it sorts these sub-paths to explore the path coherency in the path space. Finally, it designs a set of weighting strategies to better compute weights of different sampling techniques. Experiments show that the algorithm produces less noisy results under the same conditions.",

keywords = "Global illumination, Markov chain Monte Carlo, Multiple importance sampling, Path reuse",

author = "Zuo Zeng and Meie Fang and Lizhuang Ma and Xiao Lin",

year = "2016",

month = sep,

day = "1",

language = "英语",

volume = "28",

pages = "1459--1464",

journal = "Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics",

issn = "1003-9775",

publisher = "Institute of Computing Technology",

number = "9",

}

TY - JOUR

T1 - Improved multiplexed metropolis light transport

AU - Zeng, Zuo

AU - Fang, Meie

AU - Ma, Lizhuang

AU - Lin, Xiao

PY - 2016/9/1

Y1 - 2016/9/1

N2 - Aiming at the problem that multiplexed Metropolis light transport (MMLT) may not promise coherency in path space and does not reuse sub-paths, this paper introduces an improved algorithm. It firstly samples and selects sub-paths to reuse paths. Secondly, it sorts these sub-paths to explore the path coherency in the path space. Finally, it designs a set of weighting strategies to better compute weights of different sampling techniques. Experiments show that the algorithm produces less noisy results under the same conditions.

AB - Aiming at the problem that multiplexed Metropolis light transport (MMLT) may not promise coherency in path space and does not reuse sub-paths, this paper introduces an improved algorithm. It firstly samples and selects sub-paths to reuse paths. Secondly, it sorts these sub-paths to explore the path coherency in the path space. Finally, it designs a set of weighting strategies to better compute weights of different sampling techniques. Experiments show that the algorithm produces less noisy results under the same conditions.

KW - Global illumination

KW - Markov chain Monte Carlo

KW - Multiple importance sampling

KW - Path reuse

UR - https://www.scopus.com/pages/publications/84985911613

M3 - 文章

AN - SCOPUS:84985911613

SN - 1003-9775

VL - 28

SP - 1459

EP - 1464

JO - Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics

JF - Jisuanji Fuzhu Sheji Yu Tuxingxue Xuebao/Journal of Computer-Aided Design and Computer Graphics

IS - 9

ER -

Improved multiplexed metropolis light transport

Abstract

Keywords

Other files and links

Fingerprint

Cite this